1794 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO.

4, OCTOBER 2004

Optimal Placement, Replacement and Sizing

of Capacitor Banks in Distorted Distribution
Networks by Genetic Algorithms
Mohammad A. S. Masoum, Marjan Ladjevardi, Akbar Jafarian, and Ewald F. Fuchs, Fellow, IEEE

Abstract—This paper presents a new genetic algorithm A. Literature Survey

(GA)–based approach for the simultaneous power quality
improvement and optimal placement and sizing of fixed capacitor Most of the reported techniques for capacitor placement as-
banks in radial distribution networks in the presence of voltage sume sinusoidal operating conditions. These methods include:
and current harmonics. The objective function includes the cost nonlinear programming [1], [2], near global methods (genetic
of power losses, energy losses and that of the capacitor banks. algorithms [3]–[9], simulated annealing [10]–[13], tabu search
Constraints include voltage limits, number/size and locations of [14]–[17]), artificial neural networks [18]–[20] and fuzzy set
installed capacitors (at each bus and the entire feeder) and the
power quality limits of standard IEEE-519. Candidate buses for theory [21]–[25]. All these approaches ignore the presence of
capacitor placement are selected based on an initial generation of voltage and current harmonics.
chromosomes. Using a proposed fitness function, a suitable com- Some of the recent publications have taken into account the
bination of objective and constraints is defined as a criterion to presence of distorted voltages for solving the capacitor place-
select (among the candidates) the most suitable buses for capacitor ment problem. These investigations include: exhaustive search
placement. A genetic algorithm computes improved generations of
chromosomes and candidate buses until the solution is obtained. [26], local variations [27], mixed integer-nonlinear program-
Simulation results for two IEEE distorted networks are presented ming [28], heuristic methods for simultaneous capacitor and
and solutions of the genetic algorithm are compared with those filter placement [29], maximum sensitivities selection [30]–[32]
of the maximum-sensitivities-selection (MSS), the maximum and fuzzy set theory [33].
sensitivities selection-local variations (MSS-LV), and the fuzzy set
algorithms. The main contribution of this paper is the computa-
tion of the near global solution, with weak dependency on initial
B. Problems Associated With Above Techniques
conditions. These techniques are fast, but they suffer from the inability to
Index Terms—Capacitor banks, genetic algorithms, harmonics, escape local optimal solutions. Simulated annealing (SA), tabu
placement, power flow and optimization, sizing. search (TS) and genetic algorithms (GAs) are three near global
optimization techniques that have demonstrated fine capabilities
for capacitor placement under sinusoidal conditions. GAs, SA
I. INTRODUCTION
and TS methods have the capability to search a near global op-

P ROPER placement and sizing of shunt capacitor banks in

distorted networks can result in reactive power compen-
sation, improved voltage regulation, power factor correction,
timal solution but the computational burden is heavy [3]–[17].

C. Justification for Using Genetic Algorithms

power/energy loss reduction, as well as power quality improve- Most traditional optimization methods move from one point
ment. Optimal capacitor bank placement is a well-researched in the decision hyperspace to another using some deterministic
subject. However, very limited attention is given to this problem rule. The problem with this is that it is likely to get stuck at
in the presence of voltage and current harmonics. a local optimum. GAs start with a diverse set (population) of
The capacitor placement problem consists of determining the potential solutions (hyperspace vectors). This allows for explo-
optimal numbers, types, locations and sizes of capacitor banks ration of many optimums in parallel, lowering the probability of
such that minimum yearly cost due to power/energy losses and getting stuck at a local optimum [34]–[36].
cost of capacitors is achieved, while the operational and power Even though GAs are probabilistic, they are not strictly
quality constraints are maintained within the required limits. random search. The stochastic operators used in the operations
on the population direct the search toward regions of the hyper-
space that are likely to have higher fitness values.
Manuscript received January 24, 2003; revised April 26, 2003. Paper no.
TPWRD-00028-2003. D. Solution Approach
M. A. S. Masoum is with the Department of Electrical and Computer Engi-
neering, Curtin University of Technology, Perth, WA 6845, Australia. This paper proposes a genetic method to formulate the ca-
M. Ladjevardi and A. Jafarian are with the Department of Electrical Engi- pacitor placement and sizing problem in the presence of voltage
neering, Iran University of Science and Technology, Tehran 16844, Iran. and current harmonics, taking into account fixed capacitors with
E. F. Fuchs is with the Electrical and Computer Engineering Department,
University of Colorado, Boulder, CO 80309 USA. a limited number of capacitor banks at each bus and the entire
Digital Object Identifier 10.1109/TPWRD.2004.835438 feeder. Operational and power quality constraints include the
0885-8977/04$20.00 © 2004 IEEE

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 MASOUM et al.: OPTIMAL PLACEMENT, REPLACEMENT AND SIZING OF CAPACITOR BANKS 1795

bounds of rms voltage, , the number/size of installed ca- III. PROBLEM FORMULATION
pacitors and harmonic parallel resonances. In this paper, the following assumptions are made:
• capacitors with fixed values ,
II. SYSTEM MODEL AT HARMONIC FREQUENCIES • presence of linear and nonlinear loads in a balanced three-
For modeling of a distribution system at fundamental and har- phase system.
monic frequencies the formulation and notations of [37], [38]
A. Constraints
are used. System solution is achieved by forcing total (funda-
mental and harmonic) mismatch active and reactive powers as Voltage constraints will be taken into account by specifying
well as mismatch active and reactive fundamental and harmonic upper (e.g., ) and lower (e.g., )
currents to zero using the Newton–Raphson method. bounds of rms voltage ( , ):
Define bus #1 to be the conventional swing bus, buses
#2 through to be the conventional linear (PQ (3)
and PV) buses, and buses #m through #n as nonlinear buses
. We assume that nonlinear load
models—representing the coupling between harmonic volt- The distortion of voltage is considered by specifying for max-
ages and currents—are given either in the frequency domain imum total harmonic distortion of voltages:
(e.g., and characteristics) or in the time domain (e.g.,
v(t) and i(t) characteristics). These models are available for
many nonlinear loads and systems such as power electronic de-
vices [37], nonlinear transformers [38], discharge lighting [39],
EHV and HVDC networks [40], [41]. The Newton–Raphson
method [37] is used to compute the correction terms by forcing
appropriate mismatches to zero (4)

Bounds for (3), (4) are specified by the IEEE-519 standard [42].
(1) Let and denote the maximum number of capaci-
tors allowed at each bus and the entire feeder, respectively. The
For the harmonic power flow analysis, we have number of capacitors on bus i is limited by:
:
(5A)

where denotes the number of fixed capacitor banks at bus i.

The number of capacitors of the entire feeder is limited by:

.. .. .. .. .. .. (5B)
. . . . . .

where SC is the set of MC candidate buses possible for capacitor

placement.

B. Objective Function
The objective function used for capacitor placement is [26],
.. (2A) [30], [33]:
.

where subvectors and submatrices are defined in [37], [38].

In the above formulation of harmonic power flow, we assume (6)
that the capacitors are shunt capacitor banks with variable reac-
tances and capacitor placement is possible for MC number of where is the energy loss cost, is the cost of fixed
buses (candidate buses). We also use the proposed formulation capacitors and is the cost corresponding to losses (e.g.,
of [26] in the harmonic power flow for the admittance of linear used capacity of the system), and
loads at harmonic frequencies: total system losses.
bus voltage vector at harmonic h [37], [38]
the highest order of harmonics considered
size of connected capacitors
(2B) saving per MW for reduction in losses (e.g.,
, [30])

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 1796 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 4, OCTOBER 2004

TABLE I
YEARLY COST OF FIXED CAPACITORS [27]

Fig. 1. Proposed chromosome structure for the genetic algorithm.

cost per MWh (e.g., , [30])

duration of load (hours/year)
set of possible shunt capacitor buses
cost per unit of fixed capacitance ( ,
Table I) Fig. 2. Proposed penalty functions used to compute fitness (8): (a) for the
Total losses can be computed using harmonic power flow out- THD
objective function, (b) for V
, (c) for , (d) for total number of capacitors
puts: of the entire feeder (U).

the number of possible compensation (candidate) buses for ca-

pacitor placement in the entire feeder. The binary numbers in-
dicate the size of the installed capacitor at the bus under consid-
(7) eration.

B. Proposed Fitness Function

where and are magnitude and phase of hth harmonic
voltage at bus i and and are magnitude and phase of hth The inverse algebraic product of proposed penalty functions
harmonic line admittance between buses i and j, respectively. is used as the fitness function to combine the objective and con-
straint functions.
IV. GENETIC ALGORITHM
(8A)
Genetic algorithms use the principle of natural evolution and
population genetics to search and arrive at a high quality near
(8B)
global solution. The required design variables are encoded into
a binary string as a set of genes corresponding to chromosomes
in biological systems. Unlike the traditional optimization tech- (8c)
niques that require one starting point, they use a set of points
as the initial conditions. Each point is called a chromosome. A
group of chromosomes are called a population. The number of where proposed penalty functions , , and
chromosomes in a population is usually selected to be between are shown in Fig. 2.
30 to 300. Each chromosome is a string of binary codes (genes)
and may contain substrings. The merit of a string is judged by C. Genetic Operators
the fitness function, which is derived from the objective function Genetic operators are the stochastic transition rules applied
and is used in successive genetic operations. During each iter- to each chromosome during each generation procedure to gen-
ative procedure (referred to as generation), a new set of strings erate a new improved population from an old one. A genetic
with improved performance is generated using three GA opera- algorithm usually consists of reproduction, crossover and muta-
tors (namely reproduction, crossover and mutation). tion operators.
• Reproduction—is a probabilistic process for selecting
A. Structure of Chromosomes two parent strings from the population of strings on the
In this paper, the chromosome structure for GA consists of basis of “roulette-wheel” mechanism, using their fitness
MC substrings of binary numbers (Fig. 1), where MC denotes values. This ensures that the expected number of times a

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 MASOUM et al.: OPTIMAL PLACEMENT, REPLACEMENT AND SIZING OF CAPACITOR BANKS 1797

string is selected is proportional to its fitness relative to Step 1) input system parameters (e.g., system topology, line
the rest of the population. Therefore, strings with higher and load specifications). Input the initial population with
fitness values have a higher probability of contributing chromosomes.
offspring. Step 2) Set initial counter and parameter values (e.g.,
• Crossover—is the process of selecting a random position and ).
in the string and swapping the characters either left or right Step 3) (Fitness Process):
of this point with another similarly partitioned string. This Step 3A: Run harmonic power flow for chromosome
random position is called the crossover point. In this paper and save outputs.
the characters to the right of a crossover point are swapped. Step 3B: Compute proposed penalty functions (Fig. 2)
The probability of parent-chromosomes crossover are as- using outputs of harmonic power flow. Compute fitness
sumed to be between 0.6 and 1.0. functions (8) for chromosome . Set .
• Mutation—is the process of random modification of a Step 3C: If go to Step 3A.
string position by changing “0” to “1” or vice versa, with Step 4) (Reproduction Process):
a small probability. It prevents complete loss of genetic Step 4A: Define total fitness as the sum of all fitness values
material through reproduction and crossover by ensuring for all chromosomes.
that the probability of searching any region in the problem Step 4B: Select a percentage of “roulette wheel” for each
space is never zero. In this paper the probability of muta- chromosome which is equal to the ratio of its fitness value
tion is assumed to be between 0.01 and 0.1. to the total fitness value.
Step 4C: Improve generation by rolling the “roulette
D. Convergence Criterion wheel” times. Select a new combination of chro-
mosomes.
The iterations (regenerations) of the proposed genetic
Step 5) (Crossover Process):
algorithm are continued until all generated chromosomes be-
Step 5A: Select a random number for mating two
come equal or the maximum number of iterations is achieved
parent chromosomes.
. Due to the randomness of GA method, the
Step 5B: If is between 0.6 and 1.0 then combine
solution tends to differ for each run, even with the same initial
the two parents, generate two offspring and go to Step 5D.
population. For this reason, it is suggested to perform multiple
Step 5C: Else, transfer the chromosome with no crossover.
runs and select the “most acceptable” solution (e.g., with most
Step 5D: Repeat steps 5A to 5C for all chromosomes.
benefits, within the permissible region of constraints).
Step 6) (Mutation Process):
Step 6A: Select a random number for mutation
E. Analysis of one chromosome.
The initial conditions for (6) (e.g., the initial compensation Step 6B: If is between 0.01 and 0.1 then apply the
buses) do not usually reside inside the permissible solution re- mutation process at a random position and go to Step 6D.
gion. Therefore, some type of criteria is required to direct the Step 6C: Else, transfer the chromosome with no mutation.
solution toward the permissible region and to select the most Step 6D: Repeat Steps 6A to 6C for all chromosomes.
appropriate buses for capacitor placement. The paper uses a fit- Step 7) (Updating Populations): Replace the old population
ness function (based on combination of F, , U and V) to with the improved population generated by Steps 2 to 6.
improve population and convergence of the solution in consec- Check all chromosomes, if there is any chromosome with
utive generations. , and , set
The inclusion of objective function and power quality con- and save it. Set .
straints will automatically eliminate all solutions generating ex- Step 8) (Convergence): If all chromosomes are the
treme values for voltages and/or currents and prevents funda- same or the maximum number of iterations is achieved
mental and harmonic parallel resonances. , then print the solution and stop, else go to
An important factor in using genetic algorithm is the proba- Step 2.
bility of crossover and mutation. If crossover probability is close
to one, crossover is performed on the whole population and the
search effect is maximized. It is a usual practice to set crossover VI. SIMULATION RESULTS
probability to a large value in the range of 0.7 to 1.0 [34]–[36]. Previous capacitor placement techniques [30]–[33] and the
Mutation operator maintains diversity in the population and proposed genetic algorithm (Fig. 2) are used to simulate two
hence used with a small probability: on the average there is only IEEE distorted networks.
one mutation per string. Thus, mutation probability is usually set
in the range of 0.1/L to 1/L ( [34]–[36]).
A. Simulation of the 6-Bus IEEE Distorted System
The MSS [30], [31], MSS-LV [32], fuzzy set [33] and
V. SOLUTION METHODOLOGY
genetic algorithms for optimal capacitor sizing and placement
The shunt capacitor placement and sizing problem in the pres- are applied to the 69 kV, 6-bus, distorted distribution system
ence of linear and nonlinear loads is solved using the proposed (Fig. 4(a)). Specifications of this system are given in [43].
genetic algorithm of Fig. 3, as follows: Simulation results are compared in Table II.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 1798 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 4, OCTOBER 2004

Fig. 3. Proposed iterative genetic algorithm for optimal placement and sizing of capacitor banks in the presence of harmonics.

The nonlinear load in Fig. 4(a) is a six-pulse rectifier with B. Simulation of the 18-Bus IEEE Distorted System
active and reactive powers of 0.2 pu (2 MW) and 0.25 pu (2.5
MVAr), respectively. Outputs of harmonic power flow show a The MSS [30], [31], MSS-LV [32], fuzzy [33] and the genetic
maximum voltage of 5.27% (column 2 of Table II). algorithm for capacitor placement and sizing are applied to the
After optimal placement and sizing of capacitor banks with 23 kV, 18-bus distorted IEEE distribution system (Fig. 4(b)).
the MSS algorithm [31], voltage decreases to 5.007% Specifications of this system are given in [44].
by allocating 0.15 pu capacitor banks and a yearly benefit of The nonlinear load in Fig. 4(b) is a six-pulse rectifier with ac-
$35 968 is achieved (column 3 of Table II). tive and reactive powers of 0.3 pu (3 MW) and 0.226 pu (2.26
Applying the MSS-LV algorithm to the system of Fig. 4(a) MVAr), respectively. Outputs of harmonic power flow show a
shows a yearly benefit of $40 098 and a voltage of 4.97% maximum voltage of 8.486% for this system (column 2
(column 4 of Table II). Optimization of system by the fuzzy of Table III). Application of MSS method [31] to Fig. 4(b) shows
set algorithm [33] indicates voltage THD of 4.99% and yearly that optimal capacitor placement results in considerable yearly
benefit of $33 758 (column 5 of Table II). benefit (e.g., $20 653 per year) but it does not limit voltage
Application of the genetic algorithm (Fig. 3 with to the desired level of 5% (column 3 of Table III). This
) to the system of Fig. 4(a) results in about is expected from the MSS method in rich harmonic configura-
the same voltage (e.g., 4.943%), and annual savings are tions, where capacitor placement is not the primary solution for
increased to $45 778 (column 6 of Table II). harmonic mitigation.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 MASOUM et al.: OPTIMAL PLACEMENT, REPLACEMENT AND SIZING OF CAPACITOR BANKS 1799

TABLE II
SIMULATION RESULTS FOR THE 6-BUS, DISTORTED IEEE SYSTEM OF FIG. (4a). PER UNIT VA = 10 MVA, PER UNIT V = 69 kV , SWING BUS voltage = 1:00 pu

TABLE III
SIMULATION RESULTS OF MSS [31], MSS-LV [32], FUZZY [33] AND GENETIC ALGORITHMS FOR THE 18-BUS, DISTORTED IEEE DISTRIBUTION SYSTEM
(FIG. (4b)). PER UNIT VA = 10 MVA, PER UNIT V = 23 kV , SWING BUS voltage = 1:05 pu

Applying the MSS-LV algorithm shows a yearly benefit of 4.88% (column 6 of Table III) and total allocated capacitance
$17 939 and voltage is limited to 4.72% (column 4 of is decreased by 16%.
Table III). Application of the fuzzy algorithm [33] to Fig. 4(b) Compared with the MSS solution, power quality control mit-
results in an acceptable voltage level and benefits igation is achieved at the expense of lower benefits. Compared
(4.899% and $14,732, column 5 of Table III). with the MSS-LV solution, greater yearly benefits are achieved
The genetic algorithm of Fig. 3 was also applied to this system while the power quality conditions are the same. Compared with
for optimal placement and sizing of capacitor banks the fuzzy solution, the same power quality conditions result
. Results show a yearly benefit of $18 949 per year (last (rows 12–15 of Table III), but greater benefits (e.g., 28% greater
row of Table III) and maximum voltage is limited to savings) are achieved. The results of Tables II and III indicate

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 1800 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 4, OCTOBER 2004

Fig. 4. Single-line diagram of (a) the 6-bus [43] and (b) the 18-bus [44] IEEE distorted system used for simulation and analysis.

that the proposed genetic algorithm captures more advantageous Simulation results for the 6 and 18 bus IEEE distorted net-
solutions than the MSS, MSS-LV and fuzzy algorithms. This works are presented and compared with the results generated
is expected because the genetic algorithm has the capability of by the MSS, MSS-LV and fuzzy set algorithms. The main con-
computing the near global solution. clusions are:
• The proposed genetic algorithm is capable of determining
C. Computing Time the near global solution. The dependency of GA solution
Unfortunately, the application of GA’s often looks like a on the initial condition is weaker than previously proposed
never-ending process. In fact, only a compromise between pop- optimization methods. However, it requires more com-
ulation size, mutation rate and so on, can lead to an adequate puting time. This is not of great concern since capacitor
algorithm that finds good solutions in a relatively short time placement is usually a planning problem.
[36]. • Compared with the MSS, MSS-LV and fuzzy set methods,
Capacitor bank placement is usually considered a planning genetic algorithm generates more suitable results (e.g.,
problem and its computing time is not of great concern. How- lower and/or greater annual benefits).
ever, the computational burden associated with the proposed ge- • Proper placement and sizing of capacitor banks results in
netic algorithm is included and compared with those of the MSS, lower system losses, greater yearly benefits, better voltage
MSS-LV and fuzzy methods as shown in row 19 of Tables II and profile and improved power quality.
III. As expected, the computing times of the proposed genetic
algorithm are longer.
REFERENCES
VII. CONCLUSIONS [1] J. J. Grainger and S. H. Lee, “Optimal size and location of shunt capac-
itors for reduction of losses in distribution feeders,” IEEE Trans. Power
A genetic algorithm is proposed for the discrete optimization App. Syst., vol. PAS-100, pp. 1105–1118, Mar. 1981.
problem of fixed shunt capacitor placement and sizing in the [2] J. J. Grainger, S. H. Lee, A. M. Byrd, and K. N. Clinard, “Proper place-
ment of capacitors for loss reduction on distribution primary feeders
presence of voltage and current harmonics. Power and energy placement,” in Proc. American Power Conf., 1980, pp. 593–603.
losses due to installed capacitors and cost of fixed capacitors are [3] K. N. Miu, H. D. Chiang, and G. Darling, “Capacitor placement, re-
used as objective function. The power quality limits of standard placement and control in large-scale distribution systems by a GA-based
two-stage algorithm,” IEEE Trans. Power Syst., vol. 12, pp. 1160–1166,
IEEE-519, maximum and minimum rms bus voltages and the Aug. 1997.
allowed number of capacitor banks (at each bus and at the entire [4] G. Boone and H. D. Chiang, “Optimal capacitor placement in distribu-
feeder) are considered as constraints. Main contributions of the tion systems by genetic algorithm,” Int. J. Electrical Power and Energy
Systems, pp. 155–162, 1993.
proposed algorithm as compared with the previous methods are: [5] S. Sundhararajan and A. Pahwa, “Optimal selection of capacitors for ra-
• Simultaneous optimal capacitor placement and power dial distribution systems using a genetic algorithm,” IEEE Trans. Power
Syst., vol. 9, pp. 1499–1507, Aug. 1994.
quality improvement. [6] M. Delfanti, G. P. Granelli, P. Marannino, and M. Montagna, “Optimal
• Defining a proper combination of objective function and capacitor placement using deterministic and genetic algorithms,” IEEE
constraints as the fitness function for improving chromo- Trans. Power Syst., vol. 15, pp. 1041–1046, Aug. 2000.
[7] G. Levitin, A. Kalyuzhny, A. Shenkman, and M. Chertkov, “Optimal
somes and selecting the most suitable buses (among can- capacitor allocation in distribution systems using a genetic algorithm and
didates) for capacitor placement. a fast energy loss computation technique,” IEEE Trans. Power Delivery,
• As a result, the proposed algorithm for capacitor sizing vol. 15, pp. 623–628, Apr. 2000.
[8] K.-H. Kim and S.-K. You, “Voltage profile improvement by capacitor
and placement escapes local optima and converges to the placement and control in unbalanced distribution systems using GA,” in
near global solution. This is not true for other proposed Proc. Power Eng. Soc. Summer Meeting, vol. 2, 1999, pp. 800–805.
algorithms (e.g., MSS, LV, MSS-LV, fuzzy). [9] T. Ghose, S. Goswami, and S. K. Basu, “Energy loss reduction in distri-
bution system by capacitor placement through combined GA-SA tech-
• Inclusion of power quality constraints in fitness functions nique,” in Proc. IEEE Region 10th Int. Conf. Global Connectivity in En-
prevents the occurrence of harmonic parallel resonance. ergy, Computer, Communication and Control, vol. 2, 1998, pp. 502–505.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.
 MASOUM et al.: OPTIMAL PLACEMENT, REPLACEMENT AND SIZING OF CAPACITOR BANKS 1801

[10] H. D. Chiang, J. C. Wang, O. Cockings, and H. D. Shin, “Optimal ca- [33] M. A. S. Masoum, A. Jafarian, M. Ladjevardi, E. F. Fuchs, and W. M.
pacitor placements in distribution systems, part I: a new formulation and Grady, “Fuzzy approach for optimal placement and sizing of capacitor
the overall problem,” IEEE Trans. Power Delivery, vol. 5, pp. 634–642, banks in the presence of harmonics,” IEEE Trans. Power Delivery, vol.
Apr. 1990. 19, pp. 822–829, Apr. 2004.
[11] , “Optimal capacitor placement in distribution systems, part II: so- [34] D. E. Goldberg, Genetic Algorithms in Search, Optimization, and Ma-
lution algorithms and numerical results,” IEEE Trans. Power Delivery, chine Learning. Reading, MA: Addison-Wesley, 1989.
vol. 5, pp. 643–649, Apr. 1990. [35] J. D. Schaffer, J. D. Caruana, L. J. Eshelman, and R. Das, “A study of
[12] H. D. Chiang, J. C. Wang, J. Tong, and G. Darling, “Optimal capacitor control parameters affecting online performance of genetic algorithms
placement and control in large-scale unbalanced distribution systems: for function optimization,” in Proc. 3rd Int. Conf. Genetic Algorithms,
system modeling and a new formulation,” in Proc. IEEE Power Eng. San Mateo, CA, 1989, pp. 51–60.
Soc., 1994, pp. 173–179. [36] L. Davis, Handbook of Genetic Algorithms. New York: Van Nostrand,
[13] , “Optimal capacitor placement and control in large-scale unbal- 1991.
anced distribution systems: system solution algorithms and numerical [37] D. Xia and G. T. Heydt, “Harmonic power flow studies, parts I and II,”
studies,” in Proc. IEEE Power Eng. Soc., 1994, pp. 180–186. IEEE Trans. Power App. Syst., vol. PAS-101, pp. 1257–1270, June 1982.
[14] H. Mori and Y. Ogita, “Parallel tabu search for capacitor placement [38] M. A. S. Masoum and E. F. Fuchs, “Transformer magnetizing current
in radial distribution systems,” in Proc. IEEE Power Eng. Soc. Winter and iron core losses in harmonic power flow,” IEEE Trans. Power De-
Meeting, vol. 4, 2000, pp. 2334–2339. livery, vol. 9, pp. 10–20, Jan. 1994.
[15] H. T. Yang, Y. C. Huang, and C. L. Huang, “Solution to capacitor place- [39] W. M. Grady and G. T. Heydt, “Prediction of power system harmonics
ment in a radial distribution system using tabu search method,” in Proc. due to gaseous discharge lighting,” IEEE Trans. Power App. Syst., vol.
Int. Conf. Energy Management and Power Delivery, , vol. 1, 1995, pp. PAS-104, pp. 558–561, Jan. 1985.
388–393. [40] A. H. El-Abiad and D. C. Tarsi, “Load flow solution of untransposed
[16] C. S. Chang and L. P. Lern, “Application of tabu search strategy in EHV networks,” in Proc. 5th Power Industry Computer Applications
solving nondifferentiable savings function for the calculation of op- Conf., Pittsburgh, PA, 1967, pp. 377–384.
timum savings due to shunt capacitor installation in a radial distribution [41] W. Song, G. T. Heydt, and W. M. Grady, “The integration of HVDC
system,” in Proc. IEEE Power Eng. Soc. Winter Meeting, vol. 4, 2000, subsystems into harmonic power flow algorithm,” IEEE Trans. Power
pp. 2323–2328. App. Syst., vol. PAS-103, pp. 1953–1961, Aug. 1984.
[17] H. Mori and Y. Ogita, “Capacitor placement using parallel tabu search [42] IEEE Recommended Practices and Requirements for Harmonic Control
in distribution systems,” in Proc. IEEE Int. Conf. Syst., Man Cybern. C, in Electric Power Systems, IEEE Std. 519-1992, 1993.
vol. 6, 1999, pp. 521–526. [43] G. T. Heydt and W. M. Grady, User Manual, HARMFLOW Version
[18] T. S. Dillon, “Artificial neural network applications to power systems 3.1. Palo-Alto, CA: Electric Power Research Institute Publication,
and their relationship to symbolic methods,” Elect. Power Energy Syst., EL-4366-CCM, 1985.
vol. 13, no. 2, pp. 66–72, 1991. [44] W. M. Grady, M. J. Samotyj, and A. H. Noyola, “The application of net-
[19] N. I. Santoso and O. T. Tan, “Neural-net based real-time control of ca- work objective functions for actively minimizing the impact of voltage
pacitors installed on distribution systems,” IEEE Trans. Power Delivery, harmonics in power systems,” IEEE Trans. Power Delivery, vol. 7, pp.
vol. 5, pp. 266–272, Jan. 1989. 1379–1386, July 1992.
[20] P. K. Dash, S. Saha, and P. K. Nanda, “Artificial neural net approach
for capacitor placement in power system,” in Proc. 1st Int. Forum on
Application of Neutral Networks to Power Systems, 1991, pp. 247–250.
[21] H. C. Chin and W. M. Lin, “Capacitor placements for distribution
systems with fuzzy algorithm,” in Proc. IEEE Region 10’s 9th Annu.
Int. Conf., Theme, Frontiers of Computer Technology, vol. 2, 1994, pp.
Mohammad A. S. Masoum received the B.S, M.S., and Ph.D. degrees in elec-
1025–1029.
trical and computer engineering in 1983, 1985, and 1991, respectively, from the
[22] H. N. Ng and M. M. A. Salama, “Fuzzy optimal capacitor sizing and
University of Colorado, Boulder.
placement,” in Proc. Can. Conf. Electrical and Computer Engineering,
Currently, he is a Faculty Member with Curtin University of Technology,
vol. 2, 1995, pp. 680–683.
Perth, Australia.
[23] C.-T. Su and C.-C. Tsi, “A new fuzzy-reasoning approach to optimum
capacitor allocation for primary distribution systems,” in Proc. IEEE Int.
Conf. Industrial Technology, 1996, pp. 237–241.
[24] H. N. Ng, M. M. A. Salama, and A. Y. Chikhani, “Capacitor placement
in distribution systems using fuzzy technique,” in Can. Conf. Electrical
Computer Engineering, vol. 2, 1996, pp. 790–793.
[25] , “Capacitor allocation by approximate reasoning, fuzzy capacitor Marjan Ladjevardi received the Bachelor’s and Master’s degree in electrical
placement,” IEEE Trans. Power Delivery, vol. 15, pp. 393–398, Jan. engineering in 1999 and 2001, respectively, from the Iran University of Science
2000. and Technology, Tehran.
[26] Y. Baghzouz, “Effects of nonlinear loads on optimal capacitor placement Her research interest includes the analysis of power systems and harmonics.
in radial feeders,” IEEE Trans. Power Delivery, vol. 6, pp. 245–251, Jan.
1991.
[27] Y. Baghzouz and S. Ertem, “Shunt capacitor sizing for radial distribution
feeders with distorted substation voltage,” IEEE Trans. Power Delivery,
vol. 5, pp. 650–657, Apr. 1990.
[28] B. Gou and A. Abur, “Optimal capacitor placement for improving power
Akbar Jafarian received the Bachelor’s and Master’s degrees in electrical engi-
quality,” in Proc. Power Eng. Meeting, vol. 1, 1999, pp. 488–492.
neering from Tabriz University, Tabriz, Iran, and the Iran University of Science
[29] C. T. Hsu, Y. H. Yan, C. S. Chen, and S. L. Her, “Optimal reactive power
and Technology, Tehran, respectively, in 1999 and 2001.
planning for distribution systems with nonlinear loads,” in Proc. IEEE
His fields of interest include electrical machine analysis, computer-aided
Region 10 Int. Conf. Computer, Communication, Control and Power En-
analysis of power systems, and power system harmonics.
gineering , 1993, pp. 330–333.
[30] Z. Q. Wu and K. L. Lo, “Optimal choice of fixed and switched capaci-
tors in radial distributions with distorted substation voltage,” Proc. Inst.
Elect. Eng., Gen., Transm. Distrib., vol. 142, no. 1, pp. 24–28, 1995.
[31] M. A. S. Masoum, M. Ladjevardi, E. F. Fuchs, and W. M. Grady,
“Optimal sizing and placement of fixed and switched capacitor banks
under nonsinusoidal operating conditions,” in Proc. IEEE Summer Ewald F. Fuchs (F’90) received the Dipl.-Ing. degree in electrical engineering
Power Meeting, July 2002, pp. 807–813. from the University of Stuttgart, Stuttgart, Germany, and the Ph.D. degree in
[32] , “Application of local variations and maximum sensitivities selec- electrical engineering from the University of Colorado, Boulder in 1967 and
tions for optimal placement of shunt capacitor banks under nonsinu- 1970, respectively.
soidal operating conditions,” in Proc. 34th Annu. North American Power Currently, he is a Professor of Electrical Engineering at the University of Col-
Symp., pp. 507–515. orado.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY SILCHAR. Downloaded on April 16,2020 at 06:26:58 UTC from IEEE Xplore. Restrictions apply.